Quick answer: how do you prepare a workforce for agentic AI?
Readiness for agentic AI is mostly a design exercise. Decide which tasks software agents will carry, which stay with people, and where work passes between them. Then teach people the specific moves at each pass: briefing, reviewing, correcting and overruling. Teams that train on tools alone rarely see much change.
What do “agentic AI” and “agent-ready” actually mean?
Agentic AI describes software that pursues a goal across several steps on its own, choosing actions as it goes, instead of waiting for a fresh prompt each time. Ask a chatbot and it replies. Give an agent a goal and it can gather the inputs, move the work forward and raise a hand when something looks off.
Once agents join a team, you have what people increasingly call a hybrid team: humans and agents dividing the work under agreed rules about who decides, who acts and who reviews. A workforce is agent-ready when its jobs, processes and skills have been shaped around that division of labour. Think of it as architecture first and curriculum second.
Why does AI adoption keep outrunning work redesign?
Tools are bought, but ways of working have to be negotiated. A purchase order clears in weeks. Redesign asks managers to loosen old habits, redraw who has the final say, and tolerate a stretch where results are less predictable.
The data shows how lopsided this is. McKinsey’s 2026 State of AI survey reports that 40% of companies above $1 billion in revenue are now scaling agents, compared with 27% the year before. Deloitte’s 2026 Human Capital Trends research, meanwhile, finds just 6% of leaders who feel they are genuinely getting somewhere on human-AI work design.
Results track that divide. McKinsey found roughly three in four high-performing AI adopters had reworked how their workflows run, against about one in four among everyone else. Microsoft’s 2026 Work Trend Index credits the surrounding environment (culture, managerial backing, how talent is developed) with 67% of reported AI impact, and individual attitude and behaviour with 32%.
If you run learning, sit with that split for a moment. Upskilling individuals, valuable as it is, reaches roughly a third of what determines success. The remainder is set by how the work around those individuals is arranged.
Beneath this sits a leadership-alignment issue. In the same Microsoft survey, only 26% of AI users felt their leaders were consistently united on AI direction. Without a shared line, people invent their own rules, and the company accumulates dozens of isolated experiments instead of one way of working.
How do you redesign work so people and AI agents can share it?
Begin with tasks rather than job titles, and make five moves in order. A job title is too blunt a unit, because one role typically blends routine coordination, discretionary judgment and relationship building.
Move one: break the role apart. Write down what the job really involves, then label each task as routine, judgment-heavy or relationship-heavy. Many teams are surprised by how much of a role turns out to be coordination.
Move two: decide who owns what. For each task choose agent, person or both. ServiceNow’s India study draws a useful line between roles that agents may largely absorb and roles where people gain an agent as a working partner.
Move three: engineer the handoffs. Most programmes skip this step. Spell out when an agent should stop and escalate, what the human inspects, and who answers for the end result. Who decides matters more than how well anyone prompts.
Move four: teach the handoff moments. This is where learning design becomes concrete. People rehearse briefing an agent, inspecting what comes back, spotting mistakes and stepping in, using their own workflow as the practice ground.
Move five: measure, then adjust. Watch quality, rework, turnaround and how confident employees feel, workflow by workflow. As agents improve, ownership shifts, so treat redesign as a cycle.
There is early evidence this pays. ServiceNow’s India AI Maturity Index reports that firms which reworked workflows around AI saw productivity rise by 63%, a point in favour of starting with the workflow rather than the software.
Which skills matter most when people oversee AI agents?
Think of the capabilities a capable team lead needs: dividing work into steps, giving a clear brief, inspecting results, dealing with the odd case and knowing which data stays private. The World Economic Forum anticipates that 39% of core worker skills will shift by 2030 (as cited by Gloat), which makes these habits to rehearse regularly and not boxes to tick once.
Verification comes first. When agents produce the first draft, the scarce skill is knowing whether it is right. ServiceNow’s team adds that people should interrogate the data and process behind an AI answer, not just proofread the answer.
Exceptions come next. Agents shine on the well-trodden path and stumble at the edges. The employee who senses that a case is off, and knows the next move, becomes the key link in the chain.
Delegation is the third. Briefing an agent with context, boundaries and a clear finish line resembles onboarding a new colleague more than operating software, and few training catalogues cover it yet.
What does this shift mean for learning leaders in India’s IT services sector?
For learning leaders in IT services, this means working at the center of business redesign. ServiceNow and Pearson’s 2025 study projects that agentic AI will reshape 10.35 million roles in India by 2030 and add over 3 million tech jobs. Manufacturing (8 million roles), retail (7.6 million) and education (2.5 million) face the heaviest change, and IT services companies will be among those building and operating agents for them.
The study also shows how unready many firms feel: 26% of Indian organisations could not say which skills they will need, and 30% ranked data security as their leading worry. A learning plan drawn up without that clarity goes stale fast, so it should trail the workflow redesign rather than lead it.
Three habits mark the organisations likely to adapt well. They tie each programme to one concrete workflow so lessons are applied within days. They begin with managers, since managers set the norms for delegating and reviewing. And they make checking AI output a baseline skill for everyone, not a specialist option.
What should learning infrastructure do for an agent-ready workforce?
A platform cannot redesign work for you. Once redesign begins, though, the learning infrastructure determines whether new skills travel across the organization. Four demands stand out.
It must tie skills to the reshaped roles and keep that mapping fresh. It must offer realistic practice for handoffs, reviews and exceptions rather than passive content. It must tailor each path, since a payroll analyst and a systems administrator supervise agents differently. And it must tell learning leaders whether behaviour at work is actually changing.
Infosys Wingspan works in this space. It as an AI-first learning platform that adapts to each learner and helps enterprises narrow skill gaps. Several features line up with the four demands. Navigator builds role-specific learning paths, hands-on playgrounds offer a place to practise technical skills, and the Learning Analytics module shows program managers how content gets used.
For agentic AI, the useful move is to make the redesigned workflow the unit of learning. Name the handoff skills for a role, build the path around them, give learners a safe place to rehearse those moments, and read the analytics to see where people still stumble. The platform delivers the learning, and the workflow redesign decides its content.
The takeaway: readiness starts with how work is designed
Companies that benefit most from agents will not be those with the longest tool list. They will be the ones that settled, task by task, who owns what, and then trained people for the handoffs. Learning leaders are well placed to drive that conversation, because they know how capability gets built.
Start small. Pick one busy workflow, map its tasks, agree ownership, and train the people at the handoffs. See what shifts, then do it again.
Sources
AI Workforce Trends 2026 (Q4 Update), Gloat: summarises McKinsey State of AI 2026, Deloitte 2026 Global Human Capital Trends, Microsoft 2026 Work Trend Index and the WEF Future of Jobs Report 2025. Verify each figure against the primary report before publishing.
India’s leap to AI maturity, ServiceNow AI Skills Research 2025 with Pearson: a 2025 report, so state the year when citing it.